A robust neural network based pulse radar detection for weak signals

In this paper we develop a neural network capable of detecting targets with weak echoes in a pulse radar. This is possible as the network is designed as a pattern classifier (in contrast to earlier work) together with a suitable but simple pre-processing of the returns. We demonstrate through simulations that such a network exhibits better range resolution and noise tolerance when compared to previous work based on neural networks. In addition, we examine the Doppler tolerance and the overall robustness of the trained network.